1 3

Med Biol Eng Comput (2014) 52:885894DOI 10.1007/s11517-014-1184-4

ORIGINAL ARTICLE

Using a lowamplitude RF pulse at echo time (LARFET) for device localization in MRI

Murat Tmer Baykal Sarioglu Senol Mutlu Yekta Ulgen Arda Yalcinkaya Cengizhan Ozturk

Received: 30 December 2013 / Accepted: 13 August 2014 / Published online: 31 August 2014 International Federation for Medical and Biological Engineering 2014

1 Introduction

MRI as a real-time imaging modality for use during inter-ventional procedures is an active research area. Although the spatial resolution with MRI is not close to that of X-ray methods, more anatomical detail can be visualized because of increased soft-tissue contrast without exposure to ion-izing radiation. Temporal resolution in MRI is also lower than that of X-ray techniques but demonstrably sufficient to guide catheterization in humans [32]. MRI is also attractive for flexible slice positioning and its ability to display differ-ent tissue contrasts, motion, and flow. Techniques for cath-eter tracking in MRI can be grouped into three main cate-gories: passive, semi-active, and active methods [10]. Most interventions rely on tracking the interventional devices tip (tip tracking) either in real-time images or in the coor-dinate system of the MR scanner [2, 3, 9, 22]. Beyond, defining the device tip, visualization of the interventional devices shaft [19, 29] may be critical in many clinical sce-narios, especially when flexible devices, such as catheters, are navigated. But even with shaft visibility, tracking of tip is crucial for most clinical procedures.

Passive tracking involves the use of contrast agents or paramagnetic materials to enhance the contrast between the catheter and the surrounding anatomy [25, 36]. Despite their simplicity in implementation, the passive methods pri-marily suffer from low contrast-to-noise ratios, and some-times the appearance of the marker is indistinguishable from image artifacts [31]. Additionally, the catheter must be manipulated within the imaging plane since no posi-tional coordinates are generated for automatic scan plane prescription.

Semi-active techniques are based on local resonant cir-cuits without external connections. Through inductive cou-pling to the transmit coil, these resonant circuits provide

Abstract We describe a new method for frequency down-conversion of MR signals acquired with the radio-fre-quency projections method for device localization. A low-amplitude, off-center RF pulse applied simultaneously with the echo signal is utilized as the reference for frequency down-conversion. Because of the low-amplitude and large offset from the Larmor frequency, the RF pulse minimally interfered with magnetic resonance of protons. We con-ducted an experiment with the coil placed at different posi-tions to verify this concept. The down-converted signal was transformed into optical signal and transmitted via fiber-optic cable to a receiver unit placed outside the scanner room. The position of the coil could then be determined by the frequency analysis of this down-converted signal and superimposed on previously acquired MR images for com-parison. Because of minimal positional errors (0.8 mm), this new device localization method may be adequate for most interventional MRI applications.

Keywords Interventional MRI Device localization Catheter tracking Frequency down-conversion Self-mixing

M. Tmer (*) Y. Ulgen C. Ozturk Institute of Biomedical Engineering, Bogazii University, Istanbul, Turkeye-mail: murat.tumer@boun.edu.tr

B. Sarioglu Department of Electrical and Electronic Engineering, Bilgi University, Istanbul, Turkey

S. Mutlu A. Yalcinkaya Department of Electrical and Electronic Engineering, Bogazii University, Istanbul, Turkey

886 Med Biol Eng Comput (2014) 52:885894

1 3

flip angle amplification during RF transmission and signal amplification during reception, resulting in local signal enhancement [7, 8, 28, 29, 39]. In these techniques, the coil, and consequently, the catheter tip can be localized by projecting the volume onto orthogonal planes. The signal peak in the projections corresponds to the location of the resonant coil.

In active catheter tracking, the microcoil is directly con-nected to the MRI system and is positioned at the distal end of the catheter for localization [12, 18, 21]. The peaks in the frequency spectrum of the acquired MR signal cor-respond to the physical location [11]. It is also useful for signals from device channels to be displayed in color for better visualization [16, 26]. Since this technique provides the location directly, it can also be used in automated cath-eter tracking, where the imaging plane and imaging param-eters are adjusted by the tracking results [5, 38]. Orienta-tion information can also be obtained using more than one coil [17] or the shaft of the guidewire as a loopless antenna [19].

The major problem with active techniques is RF heat-ing due to long conducting wires [24, 41]. Different meth-ods have been proposed to reduce heating [6, 37, 40] or for heating-controlled operation [34] without avoiding the con-ducting wires. Additionally, fully optical systems were pro-posed that replace these conducting wires with inherently RF-safe optical fibers, eliminating the risk of RF heating [13, 14]. In these systems, the MR signal is transmitted with all amplitude and phase information, allowing high-resolution imaging besides projection-based tip tracking. In this case, the SNR suffers from the electro-optical signal conversion distally (and opto-electrical conversion proxi-mally) at this high frequency.

Amplifying [27] and frequency down-converting [1, 33] the MR signal at the catheter tip could minimize sig-nal losses and thus provide higher SNR. The mixing signal

could be provided from an outside generator, necessitating an additional cable to transmit this signal to the catheter tip. This would increase the complexity at the distal end. A second alternative is an oscillator placed at the catheter tip next to the low-noise amplifier (LNA), and mixer [1]. But for on-chip oscillators a decrease in accuracy is expected over time due to temperature or bias voltage variations.

In this work, we propose to provide the reference sig-nal for frequency down-conversion from the MRI scanners own transmitter. Since no frequency drift for the transmitter is expected over a specific imaging sequence interval, it is a very reliable source of signal. A low-amplitude RF pulse applied at echo time (LARFET) is acquired by the antenna (microcoil) and used as the reliable and precise reference to down-convert the echo signal. The frequency of LARFET is defined as an offset with respect to the center frequency of the scanner, making the technique immune to main field drifts. Following down-conversion, the electrical signal is converted into optical signal for MR-safe transmission. This is also critical to protect the circuitry of the distal unit and receiver from currents that would be induced on the outer surface of the cable.

2 Materials and methods

Three pieces of a CT dose index (CTDI) phantom were placed in 12.5 cm intervals as shown in Fig. 1. On each piece, there are 9 holes in a 3 3 placement with 10.61 cm grid interval, resulting in a 21.22 21.22 25 cm3 FoV to investigate and twenty-seven positions in 3-D coordinate system. A circuit consisting of an LNA, mixer, and optical transmitter was connected on a 4.3 6.3 cm printed cir-cuit board (PCB). The circuitry, which has a total current consumption of 24 mA, was powered from an MR-compat-ible battery (MP-174565, 4.8 Ah, Saft Batteries, Bagnolet,

Fig. 1 The locations for the experiment were selected from the holes already existing on a CTDI phantom consisting of 3 equal pieces. On each of these pieces, there are 9 holes which have a diameter of 1.3 cm and are drilled in 10.61 cm intervals (1/2 of the length of the 15 cm diagonal). They were numbered starting from the lower left cor-ner (readers view). So the holes on the first, second, and third pieces had the numbers 1119, 2129, and 3139, respectively

5

Anterior

Posterior

LeftRight

1 2 3

4 6

97 8

1

2

3

Head

Foot

887Med Biol Eng Comput (2014) 52:885894

1 3

France). A microcoil ( = 2 mm, l = 3 mm, and N = 10 turn) was used to acquire the signals. It was immersed in a saline-filled and sealed glass tube (Fig. 2a), which fits into the holes very tightly. The 3 3 grid was aligned parallel to the patient table using a level, and plastic rods (Fig. 2b). For each spatial location, an image was first acquired, in which the actual location of the microcoil was detected as a hyper-intense spot, a result of elevated RF flip angle at this location, i.e. using semi-active tracking. Following this, the location was determined using the new active method, which will be elaborated on in a subsequent section, and compared to the semi-active localization.

The instrumentation consisted of two units intercon-nected by a fiber-optic cable. The distal portion (Fig. 2a) was placed inside the scanner (Siemens Trio, Erlangen,

Germany) to acquire the MR signals and the concurrent LARFET signal. The output of this unit was an optical sig-nal and transmitted to an external unit located outside the MR room. The external unit reconverted the optical signal into electrical, which was then digitized and analyzed.

2.1 Positional localization MRI pulse sequence

The applied pulse sequence is illustrated in Fig. 3. It starts with a nonselective RF pulse applied by the body coil to excite all the spins. The amplitude of the RF pulse is a bal-ance between obtaining sufficient tracking SNR, which increases as one approaches the Ernst angle for a given repeat time (TR), and the desire for high tracking frame rate, which requires the use of short TRs. An optimal bal-ance was found to be 5 for a TR of 8 ms. After this nonse-lective RF pulse, a gradient echo is obtained by applying a readout gradient along one axis.

The angular frequency f0 of the proton magnetic moment vector that is exposed to a magnetic field B0 is given by

where is the gyromagnetic ratio whose value for 1H is 2.68 108 rad s1 T1 (/2 42.58 MHz/T). If, how-ever, a gradient G is applied in addition to the static mag-netic field, the local frequency is modified:

where x is the location of the protons with respect to the isocenter. Since , G and B0 are all constants, the preces-sion frequency of the spins is a linear function of their position x. Thus, for nonzero G, by measuring fMR, the fre-quency of the MR signal and the location of the acquisition point can be determined. Therefore, the frequency of the echo signal is a function of the coils spatial location and the magnitude of the readout gradient; hence, the location information is frequency encoded. The difference between this sequence and the conventional projection-based track-ing method (the Dumoulin method) was the inclusion of an extra, low-amplitude RF pulse (LARFET), which was applied by the body coil during signal acquisition. It served as the reference signal for frequency down-conversion. Its frequency was selected as f0 300 kHz, with a large off-set from the isocenter and outside of the field of view for unambiguous localization. The LARFET pulse amplitude was set to 0.3 V (corresponding to 2.6 dBm transmission power and 0.1 flip angle), which was still detectable by the microcoil inside the bore. The advantage of using the low-amplitude and large frequency offset of the LARFET pulse was the minimal perturbation of the MRI spins. Because of the large offset, LARFET did not excite the spins.

(1)f0 = 2pi B0,

(2)fMR = f0 + 2pi G x,

Coax cableCoil

A

Fiber- optic cable

PCB

Saline filled tube

Plastic rods

MR-compatible

battery

Level

Circuit (mixer,

LNA, etc.)

SealWood blocks

B

Fig. 2 a The distal unit. The coil was immersed in a saline-filled and sealed glass tube. It was connected to the PCB via a 7-cm coax cable and an SMA connector. This configuration helped keep the coil out-side the no-signal region shielded by the ground plane of the PCB. All stages were connected from discrete components, and at the last stage, the electrical signal was converted to optical by the LED (HFBR-1518). b The part of the experimental setup that was placed inside the scanner. Three pieces that make up a CTDI phantom were fixed using a homemade wooden framework such that the distance between them is 12.5 cm. A level put on plastic rods which were inserted into holes ensured the parallelism of the grid with respect to the patient table

888 Med Biol Eng Comput (2014) 52:885894

1 3

However, they were subject to BlochSiegert shift since an off-resonance RF field was applied [4, 30]. The phase shift of a spin can be calculated using the formula [30].

where all the frequencies are defined with respect to the rota-tional frame. Therefore, by putting RF = 2 300 kHz and B1 = 0.1/3 s 0.58 rad/s into Eq. 3, one can obtain BS 1.54 107 rad/s or fBS 2.5 108 Hz, which is negligible.

The measurements might be affected by some resonant offset conditions. Magnetic field inhomogeneities and local susceptibility gradients may also introduce similar errors. These errors are corrected by obtaining two echoes with opposite polarity gradients [11]. In this method, the loca-tion was the arithmetic mean of the independently com-puted locations from the two echoes. This procedure is repeated for each axis. Hence, four acquisitions are neces-sary for 2-D and six for 3-D localization.

The duration of the nonselective RF pulse is 0.5 ms, and the signal acquisition (readout) time is 3 ms. The readout gradient is extended to 6 ms to dephase the spins and pre-pare them for the next projection block; hence, one pro-jection block has an 8-ms duration in total. As a result, the duration of one 3-D complete projection block is approxi-mately 50 ms. This can also be reduced to 32 ms using Had-amard encoding [11] instead of the 6 excitation scheme.

2.2 Frequency encoding gradients

During the echo, together with LARFET, a frequency encoding (readout) gradient with a bandwidth of 1 kHz/

(3)BS = ( B1)2

2RF,

cm is applied. The amplitude of the gradient is determined from the following equation:

For BW = 1 kHz/cm, G is 2.349 mT/m. The opposite polarity dephasing gradient pulse applied before the read-out was rectangularly shaped with the highest possible slew rate to obtain the shortest echo time and thus the optimal spatial localization.

2.3 The distal unit

The block diagram of the distal unit is shown in Fig. 4. The microcoil has a diameter of 2 mm with an air inductance of 130 nH. It acts as an antenna to collect the MR signal together with the LARFET reference signal. The frequencies carried in the signal are labeled at each node in Fig. 4. Pas-sive detuning with crossed Schottky diodes (BAS70-04W, Infineon, Neubiberg, Germany) protects the input of the LNA from large induced voltages in the coil, although the flip angle amplification effect is not completely eliminated.

The composite signal is then fed into the LNA through a noise-matching LC network. The LNA consists of two stages: the first in common emitter (CE) configuration for low-noise operation and the second in common emittercommon base (CE-CB) cascade configuration, for its high stability. All stages use RF NPN transistors (BFP540, Infineon, Neubiberg, Germany). The LNA has a power gain of 56 dB with a noise figure of 0.9 dB at 123.2 MHz, the center frequency of the scanner.

The mixer is a Gilbert cell (HFA3101, Intersil, Milpitas, CA, USA). The output of the LNA is connected to both

(4)G =2pi BW

RF

Gradient (X, Y, Z)

Acquire

LARFET LARFET

Fig. 3 The pulse sequence used to localize the coil. A frequency encoding gradient is applied in the direction of localization with a preceding rewinder gradient. LARFET is applied as soon as the read-out gradient reaches its flattop and signal acquisition starts. The verti-cal dotted line indicates the time point where the area of the positive

polarity gradient equalized the negative part of the gradient and cor-respondingly the echo time. To overcome the effect of the resonant offset conditions, two echoes are acquired with opposed polarity gra-dients

889Med Biol Eng Comput (2014) 52:885894

1 3

inputs in order to square the signal. Self-mixing is known to be a problem occurring in direct conversion receivers when the strong local oscillator signal leaks into the other input of the mixer. When the DC offset at the mixer out-put is large, it might saturate the mixer and corrupt the sig-nal. In the current application, self-mixing was not an issue since it was deliberately used to extract the necessary sig-nal fMR fLA from the two-tone input.

Mixer output is further amplified and band-pass filtered (fC1 = 10 kHz, fC2 = 1 MHz, and AV 1000) to separate out the component at fMR fLA. Because of the high amplifica-tion in this stage, the signal is saturated and a square wave is obtained at the output, suitable for optical transmission. The signal is converted into optical signal by the HFBR-1528 (Avago Technologies, San Jose, CA, USA) and transmitted to the external unit outside the scanner room.

2.4 The external unit

The block diagram of the external unit. The signal is recon-verted into electrical form using an optical receiver HFBR-2528, which is the counterpart of the HFBR-1528, and finally captured using a digital oscilloscope (DPO4034B, Tektronix, Beaverton, USA). The frequency analysis is per-formed offline on a PC using Matlab (Mathworks, Natick, MA, USA) (Fig. 5).

2.5 Frequency analysis

Amplification of the weak MR signal at the catheter tip and down-conversion before transmission were crucial for the minimization of losses and ease of transmission. In our case, the down-converting signal was the LARFET that is simultaneously acquired with the frequency encoded MR signal. As shown in Fig. 4, fMR fLA was carrying the fre-quency encoded location information and separated out here for analysis.

The frequency analysis consisted of a fast Fourier trans-form (FFT). The peak in the frequency spectrum was local-ized, and the position of the coil was computed with the gradient amplitude taken into account.

fLA in this study was f0 = 300 kHz. Combining fPeak = fMR fLA with Eq. 2, the following was obtained:

where fPeak is the frequency of the down-converted signal. f0 was the frequency at the isocenter where B = B0. The exact value of f0 could differ from the specification provided by the scanner manufacturer. But experimental measurement of it was unnecessary since f0 was canceled out from the above equation. The above equation also shows the linear relation between fPeak and x. It can be rewritten for x for position calculation:

where 2/G is BW, the bandwidth of the gradient (Eq. 4). It gives f/x, the change in frequency of the received sig-nal with respect to change in distance from the isocenter as discussed in Sect. 2.2. As a result, the microcoil location was calculated using the formula

fPeak + f0 300 kHz = f0 + 2pi G x

fPeak 300 kHz = 2pi G x

(5)(fPeak 300 kHz) 2pi G

= x,

Location =fPeak 300 kHz

BW

fiber optic cable

MR signal

fMR Matching Tuning

Ref. Signal

fLALoop Antenna

Band-Pass

fMR, fLA

fMR, fLA

2 fMR,2 fLA,

fMR+fLA,fMR fLA,

DCfMR, fLA fMR fLALNALED

A

fMR,fLA

fMR fLAPassive Detuning

fMR fLA

MIXER

Fig. 4 The block diagram of the distal unit. Two signals are acquired by the loop antenna (tip coil). fMR is the frequency of the MR signal acquired, and fLA is the frequency of LARFET applied concurrently. After amplification, the signal is multiplied by itself, and at the out-put of the mixer, combinations of initial frequencies are generated.

Among them, fMR fLA is the location information carrying lower frequency signal. It is extracted by the band-pass filter. The amplifier at the last stage converts the signal into square waves and drives the LED

Fig. 5 The block diagram of the external unit. The frequency of the signal is the down-converted fMR fLA like at the output of the distal unit

890 Med Biol Eng Comput (2014) 52:885894

1 3

To determine the peak frequency more precisely, the data was lengthened in the time domain using zero pad-ding, resulting in higher resolution in the frequency spectrum. Zero padding to a power-of-two length also contributed to faster execution of FFT, the popular DFT algorithm.

3 Results

The concept verification experiment was conducted at 27 different positions in a 3 3 3 grid as shown in Fig. 1.

Figure 6 shows the frequency spectrum of a sample signal acquired from an off-center position. The spectrum is normalized to its peak value whose location on the fre-quency axis was used to determine the physical location of

the coil experimentally. For every position in the grid, six similar graphs were obtained, namely, for all three axes and both gradient polarities. For each axis, the physical coordi-nate is considered to be the arithmetical mean of the inde-pendently computed locations from opposite polarity gradi-ent signals.

In Fig. 7, the nine transversal MR images acquired for holes 3139 were added up and shown as one image. The computed locations were superimposed as a green plus sign for comparison. A good visual correlation was observed. Similar images can be obtained for any coplanar nine holes in the 3-D grid.

For numerical comparison, a second set of locations was extracted from the images by segmenting the bright spot regions and determining their centers as shown in Fig. 8. For a more precise localization, interpolation was per-formed within the region of interest containing the bright spot.

The graphs in Fig. 9 show the errors of the proposed sys-tem with the exact locations assumed to be the bright spot centers on the MR images. They are all 0.8 mm.

4 Discussion

A low-amplitude RF pulse was successfully used as a pre-cisely tuned down-converting signal in MRI device locali-zation. Since the location information is encoded in fre-quency, the accuracy of the localization is dependent on the frequency accuracy of the down-converting signal. In this

260 270 280 290 300 310 320 330 3400

0.2

0.4

0.6

0.8

1X: 310053

Frequency (kHz)

Nor

mal

ized

spe

ctru

m (a

.u.)

Fig. 6 A sample frequency spectrum obtained during the experi-ment. It has a sharp peak at 310,053 Hz corresponding to a location 100.5 mm off-center. The last digit can be ignored because it is below the precision of the system

Fig. 7 Left The nine transversal MR images acquired for locations 3139 are added up and shown as one image. The computed locations are also superimposed as green plus signs. Right Zoomed images for better visualization

891Med Biol Eng Comput (2014) 52:885894

1 3

case, it is as reliable as the MR scanners RF transmission system. In addition, it can be defined as an offset from the center frequency f0. Therefore, no check for f0 is necessary at the start or during the procedure to set the frequency of the LARFET.

There was a significant match with the bright spot region on the MR images and the experimentally determined loca-tions. A numerical comparison showed that the maximum

error of the system was around 0.8 mm, which corre-sponds to 1 pixel for a 512 512 image of 40 cm FoV. This is mainly due to the fact that imaging experiences the same amount of distortion from the magnetic field inho-mogeneities, as the proposed method. Taking the size of the coil into account ( = 2 mm, l = 3 mm), an error of 0.8 mm (1 pixel) is acceptable. This is also far better than required accuracies for many interventional procedures and

Fig. 8 Left Determination of the center of a sample bright spot. The center is designated with a plus sign. Interpolation is used for subpixel localization. Right 3-D plot of the region of interest. It shows the bright spot as a hill and its boundary as a contour line

Fig. 9 Absolute errors calcu-lated from the experimentally determined coordinates and the positions of the coil determined from the MR images. The numbers on the x axes give the hole number

892 Med Biol Eng Comput (2014) 52:885894

1 3

comparable to error ranges of similar active tracking sys-tems [15, 23]. Reducing the image resolution within the same FoV can result in a broader spatial coverage of the center pixel, and the location found by the system may fall in it rather than the neighbor pixel, thus virtually increasing the accuracy of image guidance. The accuracy with current configuration can also be improved, when necessary, by repeating the measurements and averaging at the expense of increased localization time. Since this method is highly sensitive to magnetic field inhomogeneities, for procedures like biopsy, where the absolute location would be needed, this method should be applied carefully and only near the isocenter, where the inhomogeneities and consecutively the distortions are smallest.

The precision of the system depends on the spacing in the discrete frequency domain data (f), which is the sam-pling frequency divided by the number of samples:

where L is signal recording time. In the formula, it seems possible to reduce the bin spacing in the frequency domain and accordingly to increase the spatial resolu-tion by lengthening the recording time L, namely, look-ing at more data at the expense of increased localization time. Zero padding, however, as mentioned in Sect. 2.5, can increase the data length N without really increasing the recording time. Although it does not improve spectral resolution, which is related to resolving closely spaced features in the frequency domain, it can provide a more precise localization of the peak frequency. With fS = 25 MS/s and L = 3 ms, 75 k data points were obtained per projection. By zero padding and extending this to 219 (512 k) points, a precision of approximately 50 Hz in peak detection and consecutively 0.5 mm in localization could be achieved with no increase in total localization time.

The orientation of the coil with regard to the main mag-netic field B0 can affect the signal quality and localization accuracy. Theoretically, there should be a sinusoidal rela-tionship between the signal intensity and the angle enclosed between B0 and the coils main axis, where maximum sig-nal is expected at 90 and minimum at 0. Erhart et al. [12] showed that this expectation is true. Since in this work, only the idea of using the LARFET as the reference signal for frequency down-conversion was tested, care was taken to keep the angle large. If placed parallel to B0, solenoid coils perform better than planar coils that are extensively proposed for tip tracking because of their suitability for mass production and parameter repeatability. Nevertheless, it is obvious that there will be a reduction in signal qual-ity even with solenoids. To overcome this undesired situa-tion, different coil geometries can be applied to this system.

f = f0N=

1L

,

Nonplanar coils [8] or tilted and decoupled multiple coils [20] are among the possible approaches.

Conventional active tracking systems achieve tempo-ral resolution down to 20 ms and spatial accuracies in the order of 0.30.5 mm. In this sense, the technique proposed here performs at least comparably well. The demonstrated performance together with the absence of electrical con-nections, we believe, paves the way for this technique to be directly applied to systems that are based on the combined nearby placement of the tip coil and circuitry for amplifi-cation and electro-optical signal conversion [1, 13, 14, 33, 35]. On the other hand, the frequency down-conversion not only simplifies the design of the stages after the LNA but also decreases the power necessary to drive the LED at the end, which is basically the electro-optical signal converter. The low frequency stage is designed to have a very high gain to saturate the signal at the output and convert the sinusoidal signal into a square wave. The loss of amplitude information is not important since in our technique, only the frequency is relevant. Conversion into square waves also simplified optical transmission of the signal since square waves can turn the LED on and off. With its original frequency, an unusual high gain LNA would be necessary to drive the LED in the same manner, which is not prefer-able due to stability concerns. A DC offset would be neces-sary to turn the LED on increasing the power consumption together with the power necessary to charge and discharge parasitic capacitances at the input of the LED.

In its current form, namely, without miniaturizing the distal unit and mounting it to the tip of a catheter, the sys-tem is fully optically isolated, and while it does not ade-quately address the RF heating issue, it is an excellent solu-tion to the challenges of patient isolation in which leakage currents must be limited to a total of 10 A. In this case, if fully charged, the battery can last up to 200 h.

The advantages of the technique we propose are as fol-lows: (1) potentially MR safe due to optical fiber, (2) pro-viding the device localization coordinates directly, (3) potential high SNR due to amplification at the source, (4) maintenance of high SNR during electro-optical and opto-electrical signal conversion and transmission due to frequency down-conversion. In addition, the power neces-sary to drive the LED (electro-optical signal converter) is reduced with the frequency down-conversion. Hence, the limited power provided by the optical power supply (in future IC-based designs) can be effectively used for signal amplification.

5 Conclusion

The proposed LARFET technique is successfully demon-strated for device localization in MRI. Down-conversion of

893Med Biol Eng Comput (2014) 52:885894

1 3

the MR signal right at the source reduced the complexity and simplified the design of the later stages for signal trans-mission. The accuracy and precision of the system are com-parable with other active tracking methods. Reducing nec-essary power can enable the miniaturization of this system, which could be mounted on catheter tips, for safe tracking of endovascular devices.

Acknowledgments The authors thank Umut Cindemir and Berk Camli for their efforts during experiments, Mr. Francis Payne and Dr. Can Akgn for their valuable editorial support. The experiments were conducted at National Magnetic Resonance Research Center (UMRAM) at Bilkent University, Ankara and Acbadem Kozyatag Hospital, Istanbul. This study was supported by The Scientific and Technological Research Council Of Turkey (TUBITAK, Project 111E197) and Bogazii University LifeSci Center (Ministry of Devel-opment, 2009K1200520), and EU Marie Curie Actions IRSES Pro-ject 269300 (TAHITI, Improving Therapy and Intervention through Imaging).

References

1. Anders J, Sangiorgio P, Deligianni X, Santini F, Scheer K, Boero G (2012) Integrated active tracking detector for MRI-guided interventions. Magn Reson Med 67(1):290

2. Bakker CJ, Hoogeveen RM, Weber J, van Vaals JJ, Viergever MA, Mali WP (1996) Visualization of dedicated catheters using fast scanning techniques with potential for MR-guided vascular interventions. Magn Reson Med 36(6):816

3. Bakker CJ, Smits HF, Bos C, van der Weide R, Zuiderveld KJ, van Vaals JJ, Hurtak WF, Viergever MA, Mali WP (1998) MR-guided balloon angioplasty: in vitro demonstration of the poten-tial of MRI for guiding, monitoring, and evaluating endovascular interventions. J Magn Reson Imaging 8(1):245

4. Bloch F, Siegert A (1940) Magnetic resonance for nonrotating fields. Phys Rev 57:522

5. Bock M, Volz S, Zuehlsdorff S, Umathum R, Fink C, Hallsc-heidt P, Semmler W (2004) MR-guided intravascular procedures: real-time parameter control and automated slice positioning with active tracking coils. J Magn Reson Imaging 19(5):580

6. Bottomley PA, Kumar A, Edelstein WA, Allen JM, Karmarkar PV (2010) Designing passive MRI-safe implantable conducting leads with electrodes. Med Phys 37(7):3828

7. Burl M, Coutts GA, Young IR (2005) Tuned fiducial markers to identify body locations with minimal perturbation of tissue mag-netization. Magn Reson Med 36(3):491

8. Busse H, Thormer G, Garnov N, Haase J, Kahn T, Moche M (2010) Technique for wireless position tracking of intravascular catheters: Performance evaluation in a vessel phantom. Proc Intl Soc Mag Reson Med 18:4163

9. Chung YC, Merkle EM, Lewin JS, Shonk JR, Duerk JL (1999) Fast T(2)-weighted imaging by PSIF at 0.2 T for interventional MRI. Magn Reson Med 42(2):335

10. Duerk JL, Wong EY, Lewin JS (2002) A brief review of hardware for catheter tracking in magnetic resonance imaging. MAGMA 13(3):199

11. Dumoulin CL, Souza SP, Darrow RD (1993) Real-time position monitoring of invasive devices using magnetic resonance. Magn Reson Med 29(3):411

12. Erhart P, Ladd ME, Steiner P, Heske N, Dumoulin CL, Debatin JF (1998) Tissue-independent MR tracking of invasive devices with an internal signal source. Magn Reson Med 39(2):279

13. Fandrey S, Weiss S, Mller J (2012) A novel active MR probe using a miniaturized optical link for a 1.5-T MRI scanner. Magn Reson Med 67(1):148

14. Fandrey S, Weiss S, Muller J (2008) Development of an active intravascular MR device with an optical transmission system. IEEE Trans Med Imaging 27(12):1723

15. Fritz J, U-Thainual P, Ungi T, Flammang AJ, Cho NB, Fichtinger G, Iordachita II, Carrino JA (2012) Augmented reality visualiza-tion with image overlay for MRI-guided intervention: accuracy for lumbar spinal procedures with a 1.5-T MRI system. Am J Roentgenol 198(3):W266

16. Guttman MA, Ozturk C, Raval AN, Raman VK, Dick AJ, DeSilva R, Karmarkar P, Lederman RJ, McVeigh ER (2007) Interventional cardiovascular procedures guided by real-time MR imaging: an interactive interface using multiple slices, adaptive projection modes and live 3D renderings. J Magn Reson Imaging 26(6):1429

17. Hillenbrand CM, Elgort DR, Wong EY, Reykowski A, Wacker FK, Lewin JS, Duerk JL (2004) Active device tracking and high-resolution intravascular MRI using a novel catheterbased, opposed-solenoid phased array coil. Magn Reson Med 51(4):668

18. Hurst GC, Hua J, Duerk JL, Cohen AM (1992) Intravascular (catheter) NMR receiver probe: Preliminary design analysis and application to canine iliofemoral imaging. Magn Reson Med 24(2):343

19. Kocaturk O, Saikus CE, Guttman MA, Faranesh AZ, Ratnayaka K, Ozturk C, McVeigh ER, Lederman RJ (2009) Whole shaft vis-ibility and mechanical performance for active MR catheters using copper-nitinol braided polymer tubes. J Cardiovasc Magn Reson 11:29

20. Kuehne T, Fahrig R, Butts K (2003) Pair of resonant fiducial markers for localization of endovascular catheters at all catheter orientations. J Magn Reson Imaging 17(5):620

21. Ladd ME, Zimmermann GG, McKinnon GC, von Schulthess GK, Dumoulin CL, Darrow RD, Hofmann E, Debatin JF (1998) Visualization of vascular guidewires using MR tracking. J Magn Reson Imaging 8(1):251

22. Lardo aC, McVeigh ER, Jumrussirikul P, Berger RD, Calkins H, Lima J, Halperin HR (2000) Visualization and temporal/spatial characterization of cardiac radiofrequency ablation lesions using magnetic resonance imaging. Circulation 102(6):698

23. Linte CA, Lang P, Rettmann ME, Cho DS, Holmes DR, Robb RA, Peters TM (2012) Accuracy considerations in image-guided cardiac interventions: experience and lessons learned. Int J Com-put Assist Radiol Surg 7(1):13

24. Nitz WR, Oppelt A, Renz W, Manke C, Lenhart M, Link J (2001) On the heating of linear conductive structures as guide wires and catheters in interventional MRI. J Magn Reson Imaging 13(1):105

25. Omary RA, Unal O, Koscielski DS, Frayne R, Korosec FR, Mis-tretta CA, Strother CM, Grist TM (2000) Real-Time MR Imag-ing-guided Passive Catheter Tracking with Use of Gadolinium- filled Catheters. J Vasc Interv Radiol 11(8):1079

26. Ozturk C, Guttman M, McVeigh E, Lederman RJ (2005) Mag-netic resonance imaging guided vascular interventions. Top Magn Reson Imaging 16(5):369

27. Pitsaer C, Umathum R, Homagk AK, Ozturk C, Bock M (2010) Three concepts for tuning and matching intravascular catheter coils. Proc Intl Soc Mag Reson Med 8:1852

28. Quick HH, Kuehl H, Kaiser G, Bosk S, Debatin JF, Ladd ME (2002) Inductively coupled stent antennas in MRI. Magn Reson Med 48(5):781

29. Quick HH, Zenge MO, Kuehl H, Kaiser G, Aker S, Massing S, Bosk S, Ladd ME (2005) Interventional magnetic resonance angiography with no strings attached: wireless active catheter visualization. Magn Reson Med 53(2):446

894 Med Biol Eng Comput (2014) 52:885894

1 3

30. Ramsey N (1955) Resonance transitions induced by perturbations at two or more different frequencies. Phys Rev 100(4):1191

31. Ratnayaka K, Faranesh AZ, Guttman MA, Kocaturk O, Saikus CE, Lederman RJ (2008) Interventional cardiovascular magnetic resonance: still tantalizing. J Cardiovasc Magn Reson 10:62

32. Razavi R, Hill DLG, Keevil SF, Miquel ME, Muthurangu V, Hegde S, Rhode K, Barnett M, van Vaals J, Hawkes DJ, Baker E (2003) Cardiac catheterisation guided by MRI in children and adults with congenital heart disease. Lancet 362(9399):1877

33. Sarioglu B, Aktan O, Oncu A, Mutlu S, Dundar G, Yalcinkaya AD (2012) An optically powered CMOS receiver system for intravascular magnetic resonance applications. IEEE J Emerg Sel Topics Power Electron 2(4):683

34. Sonmez M, Saikus CE, Bell JA, Franson DN, Halabi M, Faranesh AZ, Ozturk C, Lederman RJ, Kocaturk O (2012) MRI active guidewire with an embedded temperature probe and providing a distinct tip signal to enhance clinical safety. J Cardiovasc Magn Reson 14:38

35. Sun N, Liu Y, Lee H, Weissleder R, Ham D (2009) CMOS RF Biosensor Utilizing Nuclear Magnetic Resonance. IEEE J Solid-State Circ 44(5):1629

36. Unal O, Li J, Cheng W, Yu H, Strother CM (2006) MR-visible coatings for endovascular device visualization. J Magn Reson Imaging 23(5):763

37. Vernickel P, Schulz V, Weiss S, Gleich B (2005) A safe transmis-sion line for MRI. IEEE Trans Biomed Eng 52(6):1094

38. Wacker FK, Elgort D, Hillenbrand CM, Duerk JL, Lewin JS (2004) The catheter-driven MRI scanner: a new approach to intra-vascular catheter tracking and imaging-parameter adjustment for interventional MRI. Am J Roentgenol 183(2):391

39. Weiss S, Schaeffter T, Luedeke K, Leussler C, Holz D, Nehrke K, Rasche V, Sinkus R (1999) Catheter localization using a resonant fiducial marker during interactive MR fluoroscopy. Proc Intl Soc Mag Reson Med 12:1954

40. Weiss S, Vernickel P, Schaeffter T, Schulz V, Gleich B (2005) Transmission line for improved RF safety of interventional devices. Magn Reson Med 54(1):182

41. Yeung CJ, Atalar E (2001) A Greens function approach to local RF heating in interventional MRI. Med Phys 28(5):826

Using a low-amplitude RF pulse atecho time (LARFET) fordevice localization inMRIAbstract 1 Introduction2 Materials andmethods2.1 Positional localization MRI pulse sequence2.2 Frequency encoding gradients2.3 The distal unit2.4 The external unit2.5 Frequency analysis

3 Results4 Discussion5 ConclusionAcknowledgments References